Update README.md
Browse files
README.md
CHANGED
@@ -12,101 +12,11 @@ model-index:
|
|
12 |
|
13 |
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
|
14 |
|
|
|
15 |
|
16 |
-
|
17 |
|
18 |
-
|
19 |
-
```yaml
|
20 |
-
base_model: mistralai/Mistral-7B-v0.1
|
21 |
-
model_type: MistralForCausalLM
|
22 |
-
tokenizer_type: LlamaTokenizer
|
23 |
-
is_mistral_derived_model: true
|
24 |
-
|
25 |
-
load_in_8bit: false
|
26 |
-
load_in_4bit: true
|
27 |
-
strict: false
|
28 |
-
|
29 |
-
lora_fan_in_fan_out: false
|
30 |
-
data_seed: 49
|
31 |
-
seed: 49
|
32 |
-
|
33 |
-
datasets:
|
34 |
-
- path: _synth_data/alpaca_synth_queries_healed.jsonl
|
35 |
-
type: sharegpt
|
36 |
-
conversation: alpaca
|
37 |
-
dataset_prepared_path: last_run_prepared
|
38 |
-
val_set_size: 0.1
|
39 |
-
output_dir: ./qlora-alpaca-out
|
40 |
-
hub_model_id: hamel/hc-mistral-alpaca
|
41 |
-
|
42 |
-
adapter: qlora
|
43 |
-
lora_model_dir:
|
44 |
-
|
45 |
-
sequence_len: 896
|
46 |
-
sample_packing: false
|
47 |
-
pad_to_sequence_len: true
|
48 |
-
|
49 |
-
lora_r: 32
|
50 |
-
lora_alpha: 16
|
51 |
-
lora_dropout: 0.05
|
52 |
-
lora_target_linear: true
|
53 |
-
lora_fan_in_fan_out:
|
54 |
-
lora_target_modules:
|
55 |
-
- gate_proj
|
56 |
-
- down_proj
|
57 |
-
- up_proj
|
58 |
-
- q_proj
|
59 |
-
- v_proj
|
60 |
-
- k_proj
|
61 |
-
- o_proj
|
62 |
-
|
63 |
-
wandb_project: hc-axolotl-mistral
|
64 |
-
wandb_entity: hamelsmu
|
65 |
-
|
66 |
-
gradient_accumulation_steps: 4
|
67 |
-
micro_batch_size: 16
|
68 |
-
eval_batch_size: 16
|
69 |
-
num_epochs: 3
|
70 |
-
optimizer: adamw_bnb_8bit
|
71 |
-
lr_scheduler: cosine
|
72 |
-
learning_rate: 0.0002
|
73 |
-
max_grad_norm: 1.0
|
74 |
-
adam_beta2: 0.95
|
75 |
-
adam_epsilon: 0.00001
|
76 |
-
save_total_limit: 12
|
77 |
-
|
78 |
-
train_on_inputs: false
|
79 |
-
group_by_length: false
|
80 |
-
bf16: true
|
81 |
-
fp16: false
|
82 |
-
tf32: false
|
83 |
-
|
84 |
-
gradient_checkpointing: true
|
85 |
-
early_stopping_patience:
|
86 |
-
resume_from_checkpoint:
|
87 |
-
local_rank:
|
88 |
-
logging_steps: 1
|
89 |
-
xformers_attention:
|
90 |
-
flash_attention: true
|
91 |
-
|
92 |
-
loss_watchdog_threshold: 5.0
|
93 |
-
loss_watchdog_patience: 3
|
94 |
-
|
95 |
-
warmup_steps: 20
|
96 |
-
evals_per_epoch: 4
|
97 |
-
eval_table_size:
|
98 |
-
eval_table_max_new_tokens: 128
|
99 |
-
saves_per_epoch: 6
|
100 |
-
debug:
|
101 |
-
weight_decay: 0.0
|
102 |
-
fsdp:
|
103 |
-
fsdp_config:
|
104 |
-
special_tokens:
|
105 |
-
bos_token: "<s>"
|
106 |
-
eos_token: "</s>"
|
107 |
-
unk_token: "<unk>"
|
108 |
-
save_safetensors: true
|
109 |
-
```
|
110 |
|
111 |
# hc-mistral-alpaca
|
112 |
|
@@ -156,16 +66,30 @@ def prompt_tok(nlq, cols):
|
|
156 |
Finally, you can get predictions like this:
|
157 |
|
158 |
```python
|
|
|
159 |
nlq = "Exception count by exception and caller"
|
160 |
cols = ['error', 'exception.message', 'exception.type', 'exception.stacktrace', 'SampleRate', 'name', 'db.user', 'type', 'duration_ms', 'db.name', 'service.name', 'http.method', 'db.system', 'status_code', 'db.operation', 'library.name', 'process.pid', 'net.transport', 'messaging.system', 'rpc.system', 'http.target', 'db.statement', 'library.version', 'status_message', 'parent_name', 'aws.region', 'process.command', 'rpc.method', 'span.kind', 'serializer.name', 'net.peer.name', 'rpc.service', 'http.scheme', 'process.runtime.name', 'serializer.format', 'serializer.renderer', 'net.peer.port', 'process.runtime.version', 'http.status_code', 'telemetry.sdk.language', 'trace.parent_id', 'process.runtime.description', 'span.num_events', 'messaging.destination', 'net.peer.ip', 'trace.trace_id', 'telemetry.instrumentation_library', 'trace.span_id', 'span.num_links', 'meta.signal_type', 'http.route']
|
|
|
|
|
|
|
|
|
161 |
```
|
162 |
|
|
|
|
|
|
|
|
|
|
|
|
|
163 |
Alternatively, you can play with this model on Replicate: [hamelsmu/honeycomb-2](https://replicate.com/hamelsmu/honeycomb-2)
|
164 |
|
165 |
# Hosted Inference
|
166 |
|
167 |
This model is hosted on Replicate: (hamelsmu/honeycomb-2)[https://replicate.com/hamelsmu/honeycomb-2], using [this config](https://github.com/hamelsmu/replicate-examples/tree/master/mistral-transformers-2).
|
168 |
|
|
|
|
|
|
|
169 |
|
170 |
### Framework versions
|
171 |
|
|
|
12 |
|
13 |
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
|
14 |
|
15 |
+
### Model Description
|
16 |
|
17 |
+
A model that can generate [Honeycomb Queries](https://www.honeycomb.io/blog/introducing-query-assistant).
|
18 |
|
19 |
+
_fine-tuned by [Hamel Husain](https://hamel.dev)_
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
# hc-mistral-alpaca
|
22 |
|
|
|
66 |
Finally, you can get predictions like this:
|
67 |
|
68 |
```python
|
69 |
+
# model inputs
|
70 |
nlq = "Exception count by exception and caller"
|
71 |
cols = ['error', 'exception.message', 'exception.type', 'exception.stacktrace', 'SampleRate', 'name', 'db.user', 'type', 'duration_ms', 'db.name', 'service.name', 'http.method', 'db.system', 'status_code', 'db.operation', 'library.name', 'process.pid', 'net.transport', 'messaging.system', 'rpc.system', 'http.target', 'db.statement', 'library.version', 'status_message', 'parent_name', 'aws.region', 'process.command', 'rpc.method', 'span.kind', 'serializer.name', 'net.peer.name', 'rpc.service', 'http.scheme', 'process.runtime.name', 'serializer.format', 'serializer.renderer', 'net.peer.port', 'process.runtime.version', 'http.status_code', 'telemetry.sdk.language', 'trace.parent_id', 'process.runtime.description', 'span.num_events', 'messaging.destination', 'net.peer.ip', 'trace.trace_id', 'telemetry.instrumentation_library', 'trace.span_id', 'span.num_links', 'meta.signal_type', 'http.route']
|
72 |
+
|
73 |
+
# print prediction
|
74 |
+
out = prompt_tok(nlq, cols)
|
75 |
+
print(nlq, '\n', out)
|
76 |
```
|
77 |
|
78 |
+
This will give you a prediction that looks like this:
|
79 |
+
|
80 |
+
```md
|
81 |
+
"{'breakdowns': ['exception.message', 'exception.type'], 'calculations': [{'op': 'COUNT'}], 'filters': [{'column': 'exception.message', 'op': 'exists'}, {'column': 'exception.type', 'op': 'exists'}], 'orders': [{'op': 'COUNT', 'order': 'descending'}], 'time_range': 7200}"
|
82 |
+
```
|
83 |
+
|
84 |
Alternatively, you can play with this model on Replicate: [hamelsmu/honeycomb-2](https://replicate.com/hamelsmu/honeycomb-2)
|
85 |
|
86 |
# Hosted Inference
|
87 |
|
88 |
This model is hosted on Replicate: (hamelsmu/honeycomb-2)[https://replicate.com/hamelsmu/honeycomb-2], using [this config](https://github.com/hamelsmu/replicate-examples/tree/master/mistral-transformers-2).
|
89 |
|
90 |
+
# Training Procedure
|
91 |
+
|
92 |
+
Used [axolotl](https://github.com/OpenAccess-AI-Collective/axolotl/tree/main), see [this config](config/axolotl_config.yml).
|
93 |
|
94 |
### Framework versions
|
95 |
|